[ntp:questions] GPS-UTC Leap Corrections

cjc yheffen at gmail.com
Tue May 5 21:36:24 UTC 2009


On May 5, 10:17 am, Rob <nom... at example.com> wrote:
> cjc <yhef... at gmail.com> wrote:
> > But for some systems, it'd be nice to have them on real UTC. They'd be
> > using
> > an NTP server giving out GPS time, but we want them to have UTC on the
> > system
> > clock.
>
> Depending on where you want to see that UTC value, you could maybe define
> a UTC timezone with a 15 second offset.
> Of course time() would still return GPS, but the displayed time in
> applications would be UTC.

I considered that, but as you elude to, it doesn't really fix the main
way this manifests itself as a problem. When piece of data with a
timestamp (these are typically database entries) moves between a
system that is on GPS to one on UTC, it goes back in time. The
databases (almost all Oracle) store the "time" fields in a timezone
independent way. Fudging a funky timezone on a system can make
the time _look_ like UTC on a system, but the underlying clock is
still really GPS and that's what gets used to generate some timezone-
independent epoch any sane application is going to use under the
hood.




More information about the questions mailing list